**** BEGIN LOGGING AT Mon Jun 11 03:00:03 2018 Jun 11 07:02:38 Using pybootchart and progile.log infromation I'm able to understand the time spent for tasks and functions. But how do I analyse the time pause/gap between two tasks? Jun 11 07:04:08 Is there any way to find out what the process where doing while the build was idle between two tasks? Jun 11 09:09:29 Hi, any chance someone here can help with a bitbake query? Jun 11 09:10:47 syd: as it is not your first time here, please have a good look at: http://www.catb.org/esr/faqs/smart-questions.html Jun 11 09:11:42 haha :) Jun 11 09:12:24 well, im trying to build 2 seperate images based on 2 mostly similar distros. Jun 11 09:12:46 they share the output of recipes properly Jun 11 09:13:05 but when things should be recompiled due to change of distro features, they dont Jun 11 09:13:23 any idea how to force/inform bitbake of this? Jun 11 09:13:30 "change of distro features"? Jun 11 09:14:00 one distro has DISTRO_FEATURES with selinux, the second does not Jun 11 09:14:39 you are properly setting the distro through local.conf, and the image by passing it as parameter to bitbake? Jun 11 09:14:51 or are you trying to have the image recipe set the distro? Jun 11 09:15:20 I set the distro myself, then use bitbake Jun 11 09:15:34 i see the DISTRO is set correctly when the respective images are built Jun 11 09:15:51 "set myself" means? Jun 11 09:15:59 export DISTRO= Jun 11 09:16:43 and have you checked that those recipes that you *think* should be rebuild are actually also seeing it? Jun 11 09:17:06 yes, when i clean and build with the different DISTROs, the results are as expected Jun 11 09:17:24 clean and build = the seperate recipes Jun 11 09:18:52 well, my gut feeling is pretty clearly that at some point you are trying to do something that you are not expected to do in that way, but i can't put my finger on it. Jun 11 09:19:11 what might help you (applied reversely), is https://wiki.yoctoproject.org/wiki/TipsAndTricks/Understanding_what_changed_(diffsigs_etc) Jun 11 09:19:47 to check if the distro features are passed down like you think, especially *without* manually cleaning. Jun 11 09:20:56 thanks, ill look it through Jun 11 09:44:07 Hi there, Jun 11 09:45:41 I need a libcurl-gnutls.so.4 for a 3rd party binary to run. What is the yocto way to get this compiled? The curl package is already there Jun 11 10:08:38 darkdust: maybe oe-pkgdata-util can help you in some way, but basically its a) dive into the curl recipe to find if it has means to install said library b) pray that the ABIs match Jun 11 10:10:13 iirc curl has a gnutls/openssl toggle, so flip it to gnutls Jun 11 10:10:21 yeah, its here: http://cgit.openembedded.org/openembedded-core/tree/meta/recipes-support/curl/curl_7.60.0.bb?h=master#n25 Jun 11 10:28:22 New news from stackoverflow: include path using yocto bitbake Jun 11 10:29:14 thanks, I will try this Jun 11 10:45:12 just for clarification, I would try to create a bbappend file and change the EXTRA_OECONF variable. Is that what you meant? Jun 11 10:46:13 darkdust: bbappend yes but change packageconfig Jun 11 11:06:25 ok, I will look into it Jun 11 11:06:27 thanks Jun 11 11:42:08 Hi, its me again. i tried to check the differences between the stamps as LetoThe2nd suggested, but the specified commands dont produce stamp files Jun 11 11:43:03 my flow: bitbake -c cleanall busybox && export DISTRO= && bitbake -S none busybox && export DISTRO= && bitbake _S none busybox Jun 11 11:43:21 but my stamps directory for busybox stays empty.. any idea why this happens? Jun 11 12:03:44 hello, I'm building an image for which I've set NO_RECOMMENDATIONS = "1", but there are some packages only defined as RRECOMMENDS I would like to install using IMAGE_INSTALL Jun 11 12:04:19 an example is samba-pdb-tdbsam which is defined as RRECOMMENDS for samba-base Jun 11 12:05:25 but if I add samba-pdb-tdbsam to IMAGE_INSTALL I get an error saying "ERROR: Nothing RPROVIDES 'samba-pdb-tdbsam' (but ../src/recipes-core/images/core-image-standard.bb RDEPENDS on or otherwise requires it)" Jun 11 12:05:55 anyone know how I can add this package without enabling all RECOMMENDS ? Jun 11 12:07:30 rrerolle: sounds like you're not actually building samba-pdb-tsbsam Jun 11 12:10:40 rburton: well it's part of samba's modules which are split using 'do_split_packages' Jun 11 12:11:04 the package is built, I can see it in tmp/deploy/ipk/ Jun 11 12:11:22 tmp/deploy/ipk/core2-64/samba-pdb-tdbsam_4.7.6-r0.2_core2-64.ipk Jun 11 12:12:30 somehow bitbake seems to now know about this package, probably because it is part of an automated module split ? Jun 11 12:12:53 but I remember being able to install individual kernel module packages, which are probably built the same way Jun 11 12:13:11 not know* Jun 11 12:20:06 so what's probably happening is that the samba recipe doesn't announce that its doing dynamic splitting so bitbake doesn't know that samba will provide that recipe when its finished Jun 11 12:20:25 ie at build time, the image depends on samba-pdb-tdbsam Jun 11 12:20:32 nothing has PACKAGES = samba-pdb-tdbsam Jun 11 12:20:37 so bitbake says no Jun 11 12:20:57 samba needs a PACKAGES_DYNAMIC regex that matches the packages it can build, so bitbake knows that samba *might* build it Jun 11 12:21:06 even if its just samba-.* Jun 11 12:33:30 ok, thanks, i'll look into that Jun 11 12:58:33 rburton: it works fine when defining the packages in PACKAGES_DYNAMIC, thanks again for you help, I just submitted a PR: https://github.com/openembedded/meta-openembedded/pull/84 Jun 11 12:59:26 rrerolle: metao-oe doesn't take pull requests, send a patch to the lis Jun 11 12:59:30 list Jun 11 12:59:40 ok Jun 11 13:22:49 sorry for logging in and out, the company network don't allow IRC Jun 11 13:24:21 rburton: the PACKAGECONFIG variable has gnutls enabled but it does not build the lib. Are there further steps I have to do? Jun 11 13:59:06 New news from stackoverflow: QtWebEngine: ERR_SSL_PROTOCOL_ERROR on custom Yocto image Jun 11 14:21:47 When working with a build-server-wide sstate cache (as in multiple build-lanes to one common cache), are there any tools to erase the sstate cache for a particular package/recipe without the full build setup? Jun 11 14:22:37 Today I need to pick one an arbitrary build lanes and run bitbake -ccleanall Jun 11 14:33:12 sveinse: the cache entries have the recipe name in, so use find? Jun 11 14:33:36 rburton: and just delete them, right? Jun 11 14:33:40 yes Jun 11 14:34:13 I notice that -ccleanall often leaves *.siginfo behind. Is that intentional? Jun 11 14:38:23 What are the entities in sstate called? packages or recipes? Jun 11 14:39:05 sstate cache entries Jun 11 14:39:10 they're associated with recipes Jun 11 14:50:04 Hi, I have the following problem on fedora: https://bugzilla.redhat.com/show_bug.cgi?id=1540316 I fixed python-native with this commit https://github.com/python/cpython/commit/e348c8d154cf6342c79d627ebfe89dfe9de2381 do you already encounters this problem or i should submit it to ML? Jun 11 14:50:05 Bug 1540316: was not found. Jun 11 15:06:25 whats a good canonical way for a recipe to check if a certain variable is set, and abort if it isn't? Jun 11 15:07:44 is there way to configure a recipe to never store in the sstate cache? Jun 11 15:11:14 LetoThe2nd: check if its set and bb.fatal if it isn't Jun 11 15:11:40 rburton: in some task, i guess? Jun 11 15:11:47 anonymous py would work Jun 11 15:12:26 i see, thanks Jun 11 15:35:50 sveinse: maybe SSTATETASKS = "" ? Jun 11 15:51:33 * sveinse just realized that his 76G sstate cache consists of 42G with just the binary assets of the main application package -- which is essentially rebuilt on EVERY build Jun 11 15:54:22 I'm considering making a small tool for scanning and sorting the cache entries by size Jun 11 15:59:49 sstate-managment.sh is not enough? Jun 11 16:02:10 nayfe_: aha, so there is a tool... None of our projects have it thou, so I wasn't aware of it until now Jun 11 16:08:59 we've got scripting that uses atime... Jun 11 16:09:05 if the atime is more then 2 weeks ago, we prune Jun 11 16:09:21 (this of course assumes you have a filesystem w/ atime enabled) :) Jun 11 16:11:39 sveinse: FWIW, I just nuke the sstate cache every friday and let the nighly builds repopulate it over the weekend Jun 11 16:20:07 sveinse: if its rebuilding on every build, isn't that bad? Jun 11 16:29:42 New news from stackoverflow: How to set up syslog to log to both file && buffer Jun 11 17:30:46 rburton: no, because we only rebuild images whenever the application change. Or put another way: 95% of all changes to the image is Jun 11 17:31:01 application changes Jun 11 17:31:39 The oe/yocto system is fairly static Jun 11 17:36:00 JPEW: yeah, that is a good way. Full build without prior sstate cache will spend 5.5hrs, so it's definitely a weekend activity Jun 11 17:38:39 Has it been discussed putting a unique UUID or similar into the output packages for verification purposes? Or is there complete trust in repeatability in the compiler? Jun 11 17:39:58 sveinse: fair enough Jun 11 17:40:03 ERROR: alsa-tools-1.1.6-r0 do_package_qa: QA Issue: /usr/bin/hwmixvolume contained in package alsa-tools requires /usr/bin/python2, but no providers found in RDEPENDS_alsa-tools? [file-rdeps] Jun 11 17:40:08 but rdepends has python in :( Jun 11 17:40:09 E.g. If you build an image and make validation on this, approving its contents. Next time, since most packages are being read from the sstate cache, you don't really need to do validation on everything, since that's already done Jun 11 17:40:43 But if you wipe the cache, all packages must be rebuilt, and thus the need for revalidation (in theory) Jun 11 17:43:09 My thinking was, if the package contained a unique UUID, it would be easy to verify if the package has been rebuilt or reused from the cache Jun 11 17:45:42 sveinse: There is some work being done on making packages binary reproducable. I don't know how they verify that Jun 11 18:00:02 New news from stackoverflow: QtWebEngine: ERR_SSL_PROTOCOL_ERROR on custom Yocto image Jun 11 18:29:17 Hi, I'm trying to update my kernel version extension (accesible with uname) using my local git repository where my build system reside. this to my kernel recipe wich require recipes-kernel/linux/linux-yocto.inc Jun 11 18:29:24 GIT_VERSION ?= "" LINUX_VERSION_EXTENSION = "-custom-${GIT_VERSION}" Jun 11 18:29:46 and this task: Jun 11 18:29:46 do_preconfigure() { GIT_VERSION=$(git describe --tags --dirty) } addtask preconfigure before do_configure Jun 11 18:30:27 Anyone ever did this? On my side it always result with an empty string in GIT_VERSION variable. Jun 11 19:10:56 dkhouya: bitbake is epxanidng that ${GIT_VERSION} before it ever gets to make, using the empty bitbake variable Jun 11 19:11:07 remove the ?= "" line entirely, it'll leave it unexpanded iff it's unset Jun 11 19:15:46 Ho I see, thanks. I'll try this Jun 11 19:16:32 Should I put the LINUX_VERSION_EXTENSION = "-custom-${GIT_VERSION}" in the preconfigure task instead? Jun 11 20:08:10 halstead: I've noticed git.yoctoproject.org has been really slow recently. Know anything about that? It's pretty painful for fetching linux-yocto. Jun 11 20:19:05 net neutrality went way today. Jun 11 20:19:50 we may have to pay for priority Jun 11 20:20:19 georgem, send Halstead an email. he is out till wednesday Jun 11 20:20:58 armpit: heh. I noticed it before today but yeah who knows. Jun 11 20:21:54 Yocto project needs to switch to comcast as they are YP memebers Jun 11 20:23:32 georgem: I can look into that soon. It would be helpful to know your isp and what region you're in. Jun 11 20:23:51 halstead: Google Fiber. Kansas City area. Jun 11 20:23:53 Thanks armpit. :) Jun 11 20:25:16 georgem: thanks. Yeah that should be blazing fast. I'll find the issue on my end. Jun 11 20:27:35 * armpit kernels are not fast Jun 11 20:30:33 I am creating a distribution for a custom board. The custom board is based on an existing development board from the CPU vendor. My changes for U-Boot are not working as expected- did I use devtool incorrectly? I used 'devtool modify' to create a working copy of the vendor's u-boot and copied & modified files to add the new custom board. I did my work in 3 distinct and commitable steps, resulting in 3 patch files in my layer when Jun 11 20:30:33 I ran 'devtool finish'. When I pulled the changes into another clone for validation, it appears to be applying patches in reverse order! It is attempting to apply patch 003 but cannot find the file to modify. That file was created in patch 001. None of the files created in patch 001 are in my working directory. Jun 11 20:30:42 halstead: thanks Jun 11 22:41:57 is there a way to get openssl-native in the sysroot for the target build of a recipe ? Jun 11 23:02:58 abelloni: in native sysroot yes but target sysroot no Jun 11 23:06:27 well, I explicitely want it in the target sysroot, else, I'll have to heavily patch the makefile and I'd like to avoid that ;) Jun 11 23:14:44 I think you are approaching it wrongly if you are expecting native artifacts to be in target sysroot. Jun 11 23:15:09 I dont know what issue you are trying to solve but you should be able to stage it in sysroot-native and then reference it from there **** ENDING LOGGING AT Tue Jun 12 03:00:04 2018